Vogoo.aiLTX 2.5 is Lightricks' open-weights audio-video model, released 11 August 2026. This LTX 2.5 AI video generator runs the Pro and Fast endpoints in one browser workspace: a single prompt or still image becomes a 6 to 20 second LTX 2.5 video with synchronized audio, connected multishot scenes and up to 4K output.
LTX 2.5 video generation covers text to video, image to video with end-frame control, native multishot, synchronized audio and up to 4K output — one LTX 2.5 model instead of five separate tools.
Describe the scene, the action and the camera move, and the LTX 2.5 video model renders a finished clip with sound. A custom Gemma 4 12B text encoder tracks subjects, actions, lighting cues and camera direction through complex prompts, and Auto Duration reads the described action to pick the clip length before diffusion begins — so one LTX 2.5 prompt is enough for a usable take.
Upload a photo, a product render or a keyframe, add a line about what should happen, and LTX 2.5 animates it while holding the original subject, lighting and framing. Image to video also accepts an end frame, so you can pin exactly where the shot finishes — a start-to-end transition instead of a guess.
LTX 2.5 native multishot yields multiple connected shots that hold character, environment, lighting, voice and style across every cut. You describe a scene rather than a single shot, and the model handles the cuts internally — no stitching separate generations together, no drift between takes.
LTX 2.5 is trained as an audio-video model, so sound comes out of the same render as the picture. Turn audio on and ambience, motion sound and dialogue land on the same timeline as the visuals — no second tool, no manual sync pass, no audio multiplier in the price.
LTX 2.5 replaces plain VAE decoding with a Diffusion Video Decoder. Faces stay sharp, on-screen text stays legible, and fast motion comes through with fewer smears — the single biggest fidelity jump over LTX 2.3, visible on any frame you freeze.
The Pro endpoints run at 720p or 1080p for 6, 8 or 10 second clips with Diffusion Fidelity Rendering, which puts extra compute into complex scenes. The Fast endpoints reach 1440p and 4K at up to 50fps with clips up to 20 seconds — iterate fast in the preview, then ship the 4K cut.
The LTX 2.5 vs LTX 2.3 question comes down to a new decoder, native multishot and a stronger text encoder — the fidelity jump shows up in faces, text and fast motion.
| Capability | LTX 2.5 | LTX 2.3 |
|---|---|---|
| Video decoder | Diffusion Video Decoder (sharper faces, legible text) | VAE decoding |
| Multishot | Native multishot — connected cuts in one generation | Single-shot generation |
| Text encoder | Gemma 4 12B for complex multi-subject prompts | Previous-generation encoder |
| Audio | Synchronized audio by default | Synchronized audio by default |
| Resolution & speed | Pro 720p/1080p; Fast up to 4K and 50fps | Up to 4K output |
| Duration | Pro 6–10s; Fast up to 20s | Fast up to 20s |
| Best for | Multishot storytelling, 4K hero cuts, pro color pipelines | Short cinematic clips and rapid iteration |
LTX 2.5 is not a point release of LTX 2.3 — it replaces the decoder, adds native multishot and ships a stronger text encoder, so the fidelity gap shows up in faces, text and fast motion. On this page LTX 2.5 runs as a text-to-video and image-to-video generator.
Every LTX model before this one generated one shot at a time. LTX 2.5 is the generation that connects shots, sharpens detail and reaches 4K on the open-weights line.
The November 2024 release that put LTX on the map: a fast open-source text-to-video model from Lightricks that anyone could run locally. It set the speed bar the family still builds on.
The pro video model generation with precise control, native 4K, high frame rates and long-form creative workflows. Where most LTX comparisons start before 2.3.
The previous open-weights generation: 4K output up to 20 seconds with native audio, a new VAE for sharper details and LoRA fine-tuning support. Vogoo still runs LTX 2.3 in its video tools.
The newest LTX model, released 11 August 2026 as a 22B open-weights audio-video foundation model. Diffusion Fidelity Rendering, native multishot, a Diffusion Video Decoder and a raw pretrained checkpoint for fine-tuning — this is the model behind this page.
LTX 2.5 is built by LTX, the generative media company that spun out of Lightricks and owns the LTX model family.
LTX 2.5 fits anyone who needs short, believable motion fast — marketers, creators, product teams, filmmakers, artists and color teams.
Spin up ad variants without a shoot. Write the hook, render a 1080p LTX 2.5 video with native audio, then generate the vertical cut from the same idea and test both in the same week.
Turn a single still — a photo, a thumbnail, a piece of key art — into a 9:16 clip with sound. LTX 2.5 image to video means you can post motion without ever opening an editing timeline.
Animate a product render so the object rotates, catches light and settles with real weight. LTX 2.5 image to video with an end frame lets you pin the exact final composition for a clean loop.
Block out a shot before anyone books a crew. Describe the scene, let native multishot handle the cuts, and use the LTX 2.5 clip as a moving storyboard to argue for the setup you actually want.
Push a concept frame into motion to see whether the idea reads as an action, not just a pose. A 6 second LTX 2.5 take is cheap enough to try three directions before committing to one.
LTX 2.5 adds a native EXR workflow that reads and writes cinema-grade EXR inside ACES and DaVinci Wide Gamut, so generative edits can flow through a real color pipeline without a lossy 8-bit round trip.
The LTX 2.5 workflow stays short: describe or upload, set length, frame and resolution, then generate and download the clip.
Step 01
Type a prompt in the text-to-video tab — subjects, actions, lighting and camera moves all track through the LTX 2.5 text encoder — or switch to image to video and upload the still you want animated.
Step 02
Pick a Pro 6, 8 or 10 second clip, or a Fast clip up to 20 seconds. Choose 720p, 1080p, 1440p or 4K, select 16:9 or 9:16, and toggle native audio on or off.
Step 03
Run LTX 2.5, compare takes in the preview, then download the MP4 — all without leaving the Vogoo workspace.
Answers cover what LTX 2.5 is, whether it is free, the release date, the LTX 2.5 vs 2.3 difference, open-source licensing, downloads, audio, the MiniMax H3 comparison and multishot.
LTX 2.5 is Lightricks' latest open-weights audio-video model, released on 11 August 2026 as a 22B foundation model. It generates video with synchronized audio at up to 4K and 50fps, holds multiple connected shots in a single generation, and ships with a raw pretrained checkpoint for fine-tuning. On this page LTX 2.5 runs as a text-to-video and image-to-video generator.
Turn a prompt or a single still into an LTX 2.5 video with native audio and connected multishot scenes — up to 4K, in one Vogoo workspace.